1. Simulated Annealing Student (PhD): Umut R. ERTÜRK Lecturer : Nazlı İkizler Cinbiş 19-10-2011

2. Content:  Statistical Mechanics  NP-Complete Problems  Heuristics  Markov Chain  Monte Carlo  Simulated Annealing  Implementation Example  Strengths-Weaknesses

3. Statistical Mechanics  Ever heard Leica?  Thousands of $ But WHY?  Wetzlar was never bombed during WW2?!!! Draw a sample crystal view

4. Statistical Mechanics

5. NP Complete problems

6. Heuristics  Problem specific solutions  Two strategies  Divide and conquer  Pieces must be re-arranged after every divide  Iterative  Search/rearrange until a better configuration is found  Usually not to get stuck in local minimums, generate random configurations and save best results globally

7. Markov Chain  Next state only depends on the current state  No Memories!  Probabilistic approach  Randomness From wiki

8. Markov Chain Monte Carlo  Monte Carlo method takes place when the exact solution is not possible to find with deterministic solutions  Special Markov chain model  Which has the desire to be in a equilibrium state

9. Simulated Annealing  Using thermodynamic rules and statistical mechanics for solving NPC problems with heuristics first introduces in 1953 by Metropolis et al. (Kirkpatric, S. p.2)  Implemented in 1983 by S. Kirkpatric for solving complex problems; Travelling Salesman and Microprocessor wiring problems  Actually it is a mixture of all the subjects mentioned in this slideshow earlier.

10. Simulated Annealing  Let’s have such a decreasing function  How can we find the global minimum?  Increase x slowly and find the minimum y  How slowly? How many sampling points?

11. Simulated Annealing  What about this one?  And this one?

12. Simulated Annealing  Definition : SA is a random search technique which exploits an analogy between the way in which metal cools and freezes into a minimum energy crystalline (Busetti, F.).  For solving  combinatorial minimization problems (Teukolsky, SA)  NP complete problems (Teukolsky, SA)

13. Simulated Annealing

15.  Taken from (Teukolsky, SA, p.445)  To make use of the Metropolis algorithm for other than thermodynamic systems, one must provide the following elements: 1. A description of possible system configurations. 2. A generator of random changes in the configuration; these changes are the options” presented to the system. 3. An objective function E (analog of energy) whose minimization is the goal of the procedure. 4. 4. A control parameter T (analog of temperature) and an annealing schedule which tells how it is lowered from high to low values, e.g., after how many random changes in configuration is each downward step in T taken, and how large is that step. The meaning of “high” and “low” in this context, and the assignment of a schedule, may require physical insight and/or trial-and-error experiments.

16. Simulated Annealing  Applications  Travelling Salesman Problem  CODE!

17. Simulated Annealing StrengthsWeaknesses can deal with highly nonlinear models, chaotic and noisy data and many constraints. Since SA is a metaheuristic, a lot of choices are required to turn it into an actual algorithm. robust and general techniqueThere is a clear tradeoff between the quality of the solutions and the time required to compute them Its main advantages over other local search methods are its flexibility and its ability to approach global optimality. The tailoring work required to account for different classes of constraints and to fine- tune the parameters of the algorithm can be rather delicate The algorithm is quite versatile since it does not rely on any restrictive properties of the model The precision of the numbers used in implementation is of SA can have a significant effect upon the quality of the outcome (Busetti, F., p.4)

18. Questions? ?

19. References  Kirkpatrick, S.; Gelatt, C. D., Optimization by Simulated Annealing, 13/05/1983, SCIENCE, vol 220, p 671-680  Bertsimas D.; Tsitsiklis J., Simulated Annealing, 1993, Statistical Science vol 8 no 1, 10-15  Press H. W.; Teukolsky S.A., 2002, Numerical Recipes in C, Cambridge University Pres, p 444-456  Busetti, F., Simulated Annealing Overview, last visit: 19.10.2011, http://citeseerx.ist.psu.edu/viewdoc/download?d oi=10.1.1.66.5018&rep=rep1&type=pdf http://citeseerx.ist.psu.edu/viewdoc/download?d oi=10.1.1.66.5018&rep=rep1&type=pdf  Russel, S.; Norvig, P., Artificial Intelligence a Modern Approach, 2010, Prentice Hall, p 120-132